Abstract:
|
Randomization tests (RT), have been identified as an alternative analysis method for stepped wedge trials, as opposed to using generalized linear mixed models (GLMM) or generalized estimating equations (GEE). To our knowledge, there has been no evaluation of the study conditions under which this might be an appropriate analysis method. A simulation study is being performed to evaluate the type I and II error rates of RT versus GLMM or GEE under varying sets of stepped wedge study conditions. Evaluated conditions include each combination of the following: numbers of steps (3 or 7), clusters per step (4 or 8), participants within each cluster (20, 50, or 100 all the same size or one each of 20, 50, 75, and 100 for 4 cluster/step scenarios and one each of 20, 35, 50, 65, 80, 90, and 100 for 8 cluster/step scenarios), intra-cluster correlation (0.05 or 0.10), and effect size (none, small, or large for a given outcome). Reported results will include assessments of type I/II error rates for each of the evaluated analysis methods within each combination of factors.
|